How are virtual functions and vtable implemented?

StackOverflow https://stackoverflow.com/questions/99297

Question

We all know what virtual functions are in C++, but how are they implemented at a deep level?

Can the vtable be modified or even directly accessed at runtime?

Does the vtable exist for all classes, or only those that have at least one virtual function?

Do abstract classes simply have a NULL for the function pointer of at least one entry?

Does having a single virtual function slow down the whole class? Or only the call to the function that is virtual? And does the speed get affected if the virtual function is actually overwritten or not, or does this have no effect so long as it is virtual.

Was it helpful?

Solution

How are virtual functions implemented at a deep level?

From "Virtual Functions in C++":

Whenever a program has a virtual function declared, a v - table is constructed for the class. The v-table consists of addresses to the virtual functions for classes that contain one or more virtual functions. The object of the class containing the virtual function contains a virtual pointer that points to the base address of the virtual table in memory. Whenever there is a virtual function call, the v-table is used to resolve to the function address. An object of the class that contains one or more virtual functions contains a virtual pointer called the vptr at the very beginning of the object in the memory. Hence the size of the object in this case increases by the size of the pointer. This vptr contains the base address of the virtual table in memory. Note that virtual tables are class specific, i.e., there is only one virtual table for a class irrespective of the number of virtual functions it contains. This virtual table in turn contains the base addresses of one or more virtual functions of the class. At the time when a virtual function is called on an object, the vptr of that object provides the base address of the virtual table for that class in memory. This table is used to resolve the function call as it contains the addresses of all the virtual functions of that class. This is how dynamic binding is resolved during a virtual function call.

Can the vtable be modified or even directly accessed at runtime?

Universally, I believe the answer is "no". You could do some memory mangling to find the vtable but you still wouldn't know what the function signature looks like to call it. Anything that you would want to achieve with this ability (that the language supports) should be possible without access to the vtable directly or modifying it at runtime. Also note, the C++ language spec does not specify that vtables are required - however that is how most compilers implement virtual functions.

Does the vtable exist for all objects, or only those that have at least one virtual function?

I believe the answer here is "it depends on the implementation" since the spec doesn't require vtables in the first place. However, in practice, I believe all modern compilers only create a vtable if a class has at least 1 virtual function. There is a space overhead associated with the vtable and a time overhead associated with calling a virtual function vs a non-virtual function.

Do abstract classes simply have a NULL for the function pointer of at least one entry?

The answer is it is unspecified by the language spec so it depends on the implementation. Calling the pure virtual function results in undefined behavior if it is not defined (which it usually isn't) (ISO/IEC 14882:2003 10.4-2). In practice it does allocate a slot in the vtable for the function but does not assign an address to it. This leaves the vtable incomplete which requires the derived classes to implement the function and complete the vtable. Some implementations do simply place a NULL pointer in the vtable entry; other implementations place a pointer to a dummy method that does something similar to an assertion.

Note that an abstract class can define an implementation for a pure virtual function, but that function can only be called with a qualified-id syntax (ie., fully specifying the class in the method name, similar to calling a base class method from a derived class). This is done to provide an easy to use default implementation, while still requiring that a derived class provide an override.

Does having a single virtual function slow down the whole class or only the call to the function that is virtual?

This is getting to the edge of my knowledge, so someone please help me out here if I'm wrong!

I believe that only the functions that are virtual in the class experience the time performance hit related to calling a virtual function vs. a non-virtual function. The space overhead for the class is there either way. Note that if there is a vtable, there is only 1 per class, not one per object.

Does the speed get affected if the virtual function is actually overridden or not, or does this have no effect so long as it is virtual?

I don't believe the execution time of a virtual function that is overridden decreases compared to calling the base virtual function. However, there is an additional space overhead for the class associated with defining another vtable for the derived class vs the base class.

Additional Resources:

http://www.codersource.net/published/view/325/virtual_functions_in.aspx (via way back machine)
http://en.wikipedia.org/wiki/Virtual_table
http://www.codesourcery.com/public/cxx-abi/abi.html#vtable

OTHER TIPS

  • Can the vtable be modified or even directly accessed at runtime?

Not portably, but if you don't mind dirty tricks, sure!

WARNING: This technique is not recommended for use by children, adults under the age of 969, or small furry creatures from Alpha Centauri. Side effects may include demons which fly out of your nose, the abrupt appearence of Yog-Sothoth as a required approver on all subsequent code reviews, or the retroactive addition of IHuman::PlayPiano() to all existing instances]

In most compilers I've seen, the vtbl * is the first 4 bytes of the object, and the vtbl contents are simply an array of member pointers there (generally in the order they were declared, with the base class's first). There are of course other possible layouts, but that's what I've generally observed.

class A {
  public:
  virtual int f1() = 0;
};
class B : public A {
  public:
  virtual int f1() { return 1; }
  virtual int f2() { return 2; }
};
class C : public A {
  public:
  virtual int f1() { return -1; }
  virtual int f2() { return -2; }
};

A *x = new B;
A *y = new C;
A *z = new C;

Now to pull some shenanigans...

Changing class at runtime:

std::swap(*(void **)x, *(void **)y);
// Now x is a C, and y is a B! Hope they used the same layout of members!

Replacing a method for all instances (monkeypatching a class)

This one's a little trickier, since the vtbl itself is probably in read-only memory.

int f3(A*) { return 0; }

mprotect(*(void **)x,8,PROT_READ|PROT_WRITE|PROT_EXEC);
// Or VirtualProtect on win32; this part's very OS-specific
(*(int (***)(A *)x)[0] = f3;
// Now C::f1() returns 0 (remember we made x into a C above)
// so x->f1() and z->f1() both return 0

The latter is rather likely to make virus-checkers and the link wake up and take notice, due to the mprotect manipulations. In a process using the NX bit it may well fail.

Does having a single virtual function slow down the whole class?

Or only the call to the function that is virtual? And does the speed get affected if the virtual function is actually overwritten or not, or does this have no effect so long as it is virtual.

Having virtual functions slows down the whole class insofar as one more item of data has to be initialized, copied, … when dealing with an object of such a class. For a class with half a dozen members or so, the difference should be neglible. For a class which just contains a single char member, or no members at all, the difference might be notable.

Apart from that, it is important to note that not every call to a virtual function is a virtual function call. If you have an object of a known type, the compiler can emit code for a normal function invocation, and can even inline said function if it feels like it. It's only when you do polymorphic calls, via a pointer or reference which might point at an object of the base class or at an object of some derived class, that you need the vtable indirection and pay for it in terms of performance.

struct Foo { virtual ~Foo(); virtual int a() { return 1; } };
struct Bar: public Foo { int a() { return 2; } };
void f(Foo& arg) {
  Foo x; x.a(); // non-virtual: always calls Foo::a()
  Bar y; y.a(); // non-virtual: always calls Bar::a()
  arg.a();      // virtual: must dispatch via vtable
  Foo z = arg;  // copy constructor Foo::Foo(const Foo&) will convert to Foo
  z.a();        // non-virtual Foo::a, since z is a Foo, even if arg was not
}

The steps the hardware has to take are essentially the same, no matter whether the function is overwritten or not. The address of the vtable is read from the object, the function pointer retrieved from the appropriate slot, and the function called by pointer. In terms of actual performance, branch predictions might have some impact. So for example, if most of your objects refer to the same implementation of a given virtual function, then there is some chance that the branch predictor will correctly predict which function to call even before the pointer has been retrieved. But it doesn't matter which function is the common one: it could be most objects delegating to the non-overwritten base case, or most objects belonging to the same subclass and therefore delegating to the same overwritten case.

how are they implemented at a deep level?

I like the idea of jheriko to demonstrate this using a mock implementation. But I'd use C to implement something akin to the code above, so that the low level is more easily seen.

parent class Foo

typedef struct Foo_t Foo;   // forward declaration
struct slotsFoo {           // list all virtual functions of Foo
  const void *parentVtable; // (single) inheritance
  void (*destructor)(Foo*); // virtual destructor Foo::~Foo
  int (*a)(Foo*);           // virtual function Foo::a
};
struct Foo_t {                      // class Foo
  const struct slotsFoo* vtable;    // each instance points to vtable
};
void destructFoo(Foo* self) { }     // Foo::~Foo
int aFoo(Foo* self) { return 1; }   // Foo::a()
const struct slotsFoo vtableFoo = { // only one constant table
  0,                                // no parent class
  destructFoo,
  aFoo
};
void constructFoo(Foo* self) {      // Foo::Foo()
  self->vtable = &vtableFoo;        // object points to class vtable
}
void copyConstructFoo(Foo* self,
                      Foo* other) { // Foo::Foo(const Foo&)
  self->vtable = &vtableFoo;        // don't copy from other!
}

derived class Bar

typedef struct Bar_t {              // class Bar
  Foo base;                         // inherit all members of Foo
} Bar;
void destructBar(Bar* self) { }     // Bar::~Bar
int aBar(Bar* self) { return 2; }   // Bar::a()
const struct slotsFoo vtableBar = { // one more constant table
  &vtableFoo,                       // can dynamic_cast to Foo
  (void(*)(Foo*)) destructBar,      // must cast type to avoid errors
  (int(*)(Foo*)) aBar
};
void constructBar(Bar* self) {      // Bar::Bar()
  self->base.vtable = &vtableBar;   // point to Bar vtable
}

function f performing virtual function call

void f(Foo* arg) {                  // same functionality as above
  Foo x; constructFoo(&x); aFoo(&x);
  Bar y; constructBar(&y); aBar(&y);
  arg->vtable->a(arg);              // virtual function call
  Foo z; copyConstructFoo(&z, arg);
  aFoo(&z);
  destructFoo(&z);
  destructBar(&y);
  destructFoo(&x);
}

So you can see, a vtable is just a static block in memory, mostly containing function pointers. Every object of a polymorphic class will point to the vtable corresponding to its dynamic type. This also makes the connection between RTTI and virtual functions clearer: you can check what type a class is simply by looking at what vtable it points at. The above is simplified in many ways, like e.g. multiple inheritance, but the general concept is sound.

If arg is of type Foo* and you take arg->vtable, but is actually an object of type Bar, then you still get the correct address of the vtable. That's because the vtable is always the first element at the address of the object, no matter whether it's called vtable or base.vtable in a correctly-typed expression.

Usually with a VTable, an array of pointers to functions.

This answer has been incorporated into the Community Wiki answer

  • Do abstract classes simply have a NULL for the function pointer of at least one entry?

The answer for that is that it is unspecified - calling the pure virtual function results in undefined behavior if it is not defined (which it usually isn't) (ISO/IEC 14882:2003 10.4-2). Some implementations do simply place a NULL pointer in the vtable entry; other implementations place a pointer to a dummy method that does something similar to an assertion.

Note that an abstract class can define an implementation for a pure virtual function, but that function can only be called with a qualified-id syntax (ie., fully specifying the class in the method name, similar to calling a base class method from a derived class). This is done to provide an easy to use default implementation, while still requiring that a derived class provide an override.

You can recreate the functionality of virtual functions in C++ using function pointers as members of a class and static functions as the implementations, or using pointer to member functions and member functions for the implementations. There are only notational advantages between the two methods... in fact virtual function calls are just a notational convenience themselves. In fact inheritance is just a notational convenience... it can all be implemented without using the language features for inheritance. :)

The below is crap untested, probably buggy code, but hopefully demonstrates the idea.

e.g.

class Foo
{
protected:
 void(*)(Foo*) MyFunc;
public:
 Foo() { MyFunc = 0; }
 void ReplciatedVirtualFunctionCall()
 {
  MyFunc(*this);
 }
...
};

class Bar : public Foo
{
private:
 static void impl1(Foo* f)
 {
  ...
 }
public:
 Bar() { MyFunc = impl1; }
...
};

class Baz : public Foo
{
private:
 static void impl2(Foo* f)
 {
  ...
 }
public:
 Baz() { MyFunc = impl2; }
...
};

I'll try to make it simple :)

We all know what virtual functions are in C++, but how are they implemented at a deep level?

This is an array with pointers to functions, which are implementations of a particular virtual function. An index in this array represents particular index of a virtual function defined for a class. This includes pure virtual functions.

When a polymorphic class derives from another polymorphic class, we may have the following situations:

  • The deriving class does not add new virtual functions nor overrides any. In this case this class shares the vtable with the base class.
  • The deriving class adds and overrides virtual methods. In this case it gets its own vtable, where the added virtual functions have index starting past the last derived one.
  • Multiple polymorphic classes in the inheritance. In this case we have an index-shift between second and next bases and the index of it in the derived class

Can the vtable be modified or even directly accessed at runtime?

Not standard way - there's no API to access them. Compilers may have some extensions or private APIs to access them, but that may be only an extension.

Does the vtable exist for all classes, or only those that have at least one virtual function?

Only those that have at least one virtual function (be it even destructor) or derive at least one class that has its vtable ("is polymorphic").

Do abstract classes simply have a NULL for the function pointer of at least one entry?

That's a possible implementation, but rather not practiced. Instead there is usually a function that prints something like "pure virtual function called" and does abort(). The call to that may occur if you try to call the abstract method in the constructor or destructor.

Does having a single virtual function slow down the whole class? Or only the call to the function that is virtual? And does the speed get affected if the virtual function is actually overwritten or not, or does this have no effect so long as it is virtual.

The slowdown is only dependent on whether the call is resolved as direct call or as a virtual call. And nothing else matters. :)

If you call a virtual function through a pointer or reference to an object, then it will be always implemented as virtual call - because the compiler can never know what kind of object will be assigned to this pointer in runtime, and whether it is of a class in which this method is overridden or not. Only in two cases the compiler can resolve the call to a virtual function as a direct call:

  • If you call the method through a value (a variable or result of a function that returns a value) - in this case the compiler has no doubts what the actual class of the object is, and can "hard-resolve" it at compile time.
  • If the virtual method is declared final in the class to which you have a pointer or reference through which you call it (only in C++11). In this case compiler knows that this method cannot undergo any further overriding and it can only be the method from this class.

Note though that virtual calls have only overhead of dereferencing two pointers. Using RTTI (although only available for polymorphic classes) is slower than calling virtual methods, should you find a case to implement the same thing two such ways. For example, defining virtual bool HasHoof() { return false; } and then override only as bool Horse::HasHoof() { return true; } would provide you with ability to call if (anim->HasHoof()) that will be faster than trying if(dynamic_cast<Horse*>(anim)). This is because dynamic_cast has to walk through the class hierarchy in some cases even recursively to see if there can be built the path from the actual pointer type and the desired class type. While the virtual call is always the same - dereferencing two pointers.

Here is a runnable manual implementation of virtual table in modern C++. It has well-defined semantics, no hacks and no void*.

Note: .* and ->* are different operators than * and ->. Member function pointers work differently.

#include <iostream>
#include <vector>
#include <memory>

struct vtable; // forward declare, we need just name

class animal
{
public:
    const std::string& get_name() const { return name; }

    // these will be abstract
    bool has_tail() const;
    bool has_wings() const;
    void sound() const;

protected: // we do not want animals to be created directly
    animal(const vtable* vtable_ptr, std::string name)
    : vtable_ptr(vtable_ptr), name(std::move(name)) { }

private:
    friend vtable; // just in case for non-public methods

    const vtable* const vtable_ptr;
    std::string name;
};

class cat : public animal
{
public:
    cat(std::string name);

    // functions to bind dynamically
    bool has_tail() const { return true; }
    bool has_wings() const { return false; }
    void sound() const
    {
        std::cout << get_name() << " does meow\n"; 
    }
};

class dog : public animal
{
public:
    dog(std::string name);

    // functions to bind dynamically
    bool has_tail() const { return true; }
    bool has_wings() const { return false; }
    void sound() const
    {
        std::cout << get_name() << " does whoof\n"; 
    }
};

class parrot : public animal
{
public:
    parrot(std::string name);

    // functions to bind dynamically
    bool has_tail() const { return false; }
    bool has_wings() const { return true; }
    void sound() const
    {
        std::cout << get_name() << " does crrra\n"; 
    }
};

// now the magic - pointers to member functions!
struct vtable
{
    bool (animal::* const has_tail)() const;
    bool (animal::* const has_wings)() const;
    void (animal::* const sound)() const;

    // constructor
    vtable (
        bool (animal::* const has_tail)() const,
        bool (animal::* const has_wings)() const,
        void (animal::* const sound)() const
    ) : has_tail(has_tail), has_wings(has_wings), sound(sound) { }
};

// global vtable objects
const vtable vtable_cat(
    static_cast<bool (animal::*)() const>(&cat::has_tail),
    static_cast<bool (animal::*)() const>(&cat::has_wings),
    static_cast<void (animal::*)() const>(&cat::sound));
const vtable vtable_dog(
    static_cast<bool (animal::*)() const>(&dog::has_tail),
    static_cast<bool (animal::*)() const>(&dog::has_wings),
    static_cast<void (animal::*)() const>(&dog::sound));
const vtable vtable_parrot(
    static_cast<bool (animal::*)() const>(&parrot::has_tail),
    static_cast<bool (animal::*)() const>(&parrot::has_wings),
    static_cast<void (animal::*)() const>(&parrot::sound));

// set vtable pointers in constructors
cat::cat(std::string name) : animal(&vtable_cat, std::move(name)) { }
dog::dog(std::string name) : animal(&vtable_dog, std::move(name)) { }
parrot::parrot(std::string name) : animal(&vtable_parrot, std::move(name)) { }

// implement dynamic dispatch
bool animal::has_tail() const
{
    return (this->*(vtable_ptr->has_tail))();
}

bool animal::has_wings() const
{
    return (this->*(vtable_ptr->has_wings))();
}

void animal::sound() const
{
    (this->*(vtable_ptr->sound))();
}

int main()
{
    std::vector<std::unique_ptr<animal>> animals;
    animals.push_back(std::make_unique<cat>("grumpy"));
    animals.push_back(std::make_unique<cat>("nyan"));
    animals.push_back(std::make_unique<dog>("doge"));
    animals.push_back(std::make_unique<parrot>("party"));

    for (const auto& a : animals)
        a->sound();

    // note: destructors are not dispatched virtually
}

Each object has a vtable pointer that points to an array of member functions.

Something not mentioned here in all these answers is that in case of multiple inheritance, where the base classes all have virtual methods. The inheriting class has multiple pointers to a vmt. The result is that the size of each instance of such an object is bigger. Everybody knows that a class with virtual methods has 4 bytes extra for the vmt, but in case of multiple inheritance it is for each base class that has virtual methods times 4. 4 being the size of the pointer.

Burly's answers are correct here except for the question:

Do abstract classes simply have a NULL for the function pointer of at least one entry?

The answer is that no virtual table is created at all for abstract classes. There is no need since no objects of these classes can be created!

In other words if we have:

class B { ~B() = 0; }; // Abstract Base class
class D : public B { ~D() {} }; // Concrete Derived class

D* pD = new D();
B* pB = pD;

The vtbl pointer accessed through pB will be the vtbl of class D. This is exactly how polymorphism is implemented. That is, how D methods are accessed through pB. There is no need for a vtbl for class B.

In response to Mike's comment below...

If the B class in my description has a virtual method foo() that is not overridden by D and a virtual method bar() that is overridden, then D's vtbl will have a pointer to B's foo() and to its own bar(). There is still no vtbl created for B.

very cute proof of concept i made a bit earlier(to see if order of inheritence matters); let me know if your implementation of C++ actually rejects it(my version of gcc only gives a warning for assigning anonymous structs, but that's a bug), i'm curious.

CCPolite.h:

#ifndef CCPOLITE_H
#define CCPOLITE_H

/* the vtable or interface */
typedef struct {
    void (*Greet)(void *);
    void (*Thank)(void *);
} ICCPolite;

/**
 * the actual "object" literal as C++ sees it; public variables be here too 
 * all CPolite objects use(are instances of) this struct's structure.
 */
typedef struct {
    ICCPolite *vtbl;
} CPolite;

#endif /* CCPOLITE_H */

CCPolite_constructor.h:

/** 
 * unconventionally include me after defining OBJECT_NAME to automate
 * static(allocation-less) construction.
 *
 * note: I assume CPOLITE_H is included; since if I use anonymous structs
 *     for each object, they become incompatible and cause compile time errors
 *     when trying to do stuff like assign, or pass functions.
 *     this is similar to how you can't pass void * to windows functions that
 *         take handles; these handles use anonymous structs to make 
 *         HWND/HANDLE/HINSTANCE/void*/etc not automatically convertible, and
 *         require a cast.
 */
#ifndef OBJECT_NAME
    #error CCPolite> constructor requires object name.
#endif

CPolite OBJECT_NAME = {
    &CCPolite_Vtbl
};

/* ensure no global scope pollution */
#undef OBJECT_NAME

main.c:

#include <stdio.h>
#include "CCPolite.h"

// | A Greeter is capable of greeting; nothing else.
struct IGreeter
{
    virtual void Greet() = 0;
};

// | A Thanker is capable of thanking; nothing else.
struct IThanker
{
    virtual void Thank() = 0;
};

// | A Polite is something that implements both IGreeter and IThanker
// | Note that order of implementation DOES MATTER.
struct IPolite1 : public IGreeter, public IThanker{};
struct IPolite2 : public IThanker, public IGreeter{};

// | implementation if IPolite1; implements IGreeter BEFORE IThanker
struct CPolite1 : public IPolite1
{
    void Greet()
    {
        puts("hello!");
    }

    void Thank()
    {
        puts("thank you!");
    }
};

// | implementation if IPolite1; implements IThanker BEFORE IGreeter
struct CPolite2 : public IPolite2
{
    void Greet()
    {
        puts("hi!");
    }

    void Thank()
    {
        puts("ty!");
    }
};

// | imposter Polite's Greet implementation.
static void CCPolite_Greet(void *)
{
    puts("HI I AM C!!!!");
}

// | imposter Polite's Thank implementation.
static void CCPolite_Thank(void *)
{
    puts("THANK YOU, I AM C!!");
}

// | vtable of the imposter Polite.
ICCPolite CCPolite_Vtbl = {
    CCPolite_Thank,
    CCPolite_Greet    
};

CPolite CCPoliteObj = {
    &CCPolite_Vtbl
};

int main(int argc, char **argv)
{
    puts("\npart 1");
    CPolite1 o1;
    o1.Greet();
    o1.Thank();

    puts("\npart 2");    
    CPolite2 o2;    
    o2.Greet();
    o2.Thank();    

    puts("\npart 3");    
    CPolite1 *not1 = (CPolite1 *)&o2;
    CPolite2 *not2 = (CPolite2 *)&o1;
    not1->Greet();
    not1->Thank();
    not2->Greet();
    not2->Thank();

    puts("\npart 4");        
    CPolite1 *fake = (CPolite1 *)&CCPoliteObj;
    fake->Thank();
    fake->Greet();

    puts("\npart 5");        
    CPolite2 *fake2 = (CPolite2 *)fake;
    fake2->Thank();
    fake2->Greet();

    puts("\npart 6");        
    #define OBJECT_NAME fake3
    #include "CCPolite_constructor.h"
    fake = (CPolite1 *)&fake3;
    fake->Thank();
    fake->Greet();

    puts("\npart 7");        
    #define OBJECT_NAME fake4
    #include "CCPolite_constructor.h"
    fake2 = (CPolite2 *)&fake4;
    fake2->Thank();
    fake2->Greet();    

    return 0;
}

output:

part 1
hello!
thank you!

part 2
hi!
ty!

part 3
ty!
hi!
thank you!
hello!

part 4
HI I AM C!!!!
THANK YOU, I AM C!!

part 5
THANK YOU, I AM C!!
HI I AM C!!!!

part 6
HI I AM C!!!!
THANK YOU, I AM C!!

part 7
THANK YOU, I AM C!!
HI I AM C!!!!

note since I am never allocating my fake object, there is no need to do any destruction; destructors are automatically put at the end of scope of dynamically allocated objects to reclaim the memory of the object literal itself and the vtable pointer.

Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top