Velocity Reviews - Computer Hardware Reviews

Velocity Reviews > Newsgroups > Programming > C++ > Multiple Inheritance vs. Interface

Reply
Thread Tools

Multiple Inheritance vs. Interface

 
 
Stuart
Guest
Posts: n/a
 
      09-26-2012
On 9/23/12 "lieve again" wrote:
[snip]
> I thought maybe making the interfaces pure virtual,
> there was a way to avoid the extra vpointers and I wanted to know how.
> Then if I start adding pure virtual classes to impose the derived
> classes with some kind of features like:
> class Derived : implements Readable, Writeable, Comparable,
> Convertible ...
> regardless of the programming language, we are ending with instances
> of the derived classes having 20 bytes or more even being those
> classes with no members or empty. It is good to know.


There is one way to avoid object bloat, but you are not going to like it


The following code uses a hand-made vtable substitute. For that reason I
had to introduce a special pointer type which stores the class's ID
together with the pointer to the object. This class ID is assigned by
hand, so that this scheme will only work for class hierarchies that will
not get extended (if you want to use this for extendable class
hierarchies, you'd have to use growable look-up table instead of a fixed
array of method pointers, but then you had better use objective-C++).

Note that the base class Base in my example must not contain any virtual
methods, or else the class Derived will get bloated again. Of course,
this makes the code really awfull to look at. However, most of it may
get generated by a some clever pre-processor magic.

Since the full code may cause some shock, I'll give a short summary:

class Base {
public:
void foo () {std::cout << "Base::foo\n";}
BasePtr operator& ();
};

class Derived : public Base {
public:
void foo () {std::cout << "Derived::foo\n";}
DerivedPtr operator&();
}

Base1Ptr, Base2Ptr and DerivedPtr and those special pointer types that
allow us to do the following:

void invokeFooVirtually (const BasePtr& b) {
b.foo();
}

int main () {
Base b;
Derived d;
invokeFooVirtually(&b);
invokeFooVirtually(&d);
}

will print:

Base1::foo
Derived::foo

There is a price you have to pay: The pointer types are now twice as
large (the xxxPtr contains not only the raw pointer but also the class's
ID). Furthermore the cast from DerivedPtr to BasePtr has to call the
cast operator of DerivedPtr.



A compilable example with two base classes and some members:

#include <iostream>

class Base1Ptr {
protected:
static const int classID;
class Base1* ptr;
int objectID;
public:
Base1Ptr (class Base1* ptr, int objectID = classID)
: ptr(ptr), objectID(objectID) {}
void foo () const;
};
const int Base1Ptr::classID = 0;

class Base1 {
protected:
int base1Int;
public:
Base1 (int i) : base1Int(i / 2) {}
void foo () { std::cout << "Base1::foo with base1Int = "
<< base1Int << "\n";}
Base1Ptr operator& () {return Base1Ptr(this);}
};


class Base2Ptr {
protected:
static const int classID;
class Base2* ptr;
int objectID;
public:
Base2Ptr (class Base2* ptr, int objectID = classID)
: ptr(ptr), objectID(objectID) {}
void bar () const;
};
const int Base2Ptr::classID = 0;

class Base2 {
protected:
int base2Int;
public:
Base2(int i) : base2Int(i * 2){}
void bar () { std::cout << "Base2::bar with base2Int = "
<< base2Int << "\n";}
Base2Ptr operator* ();
};




class DerivedPtr {
class Derived* ptr;
int objectID;
public:
DerivedPtr (class Derived* ptr, int objectID)
: ptr(ptr), objectID(objectID) {}
operator Base1Ptr();
operator Base2Ptr();
};


class Derived : public Base1, public Base2 {
static const int classID;
public:
Derived (int i) : Base1(i), Base2(i){}
void foo () { std::cout << "Derived::foo with base1Int = "
<< base1Int << "and base2Int = "
<< base2Int << "\n";}
void bar () { std::cout << "Derived::bar\n";}
DerivedPtr operator&() {return DerivedPtr(this,classID);}
};

const int Derived::classID = 1; // std::max(Base1::classID,
// Base2::classID) + 1;
DerivedPtr:perator Base1Ptr() {return Base1Ptr(ptr, objectID);}
DerivedPtr:perator Base2Ptr() {return Base2Ptr(ptr, objectID);}



// Base1Ptr uses the following table to look up
// the correct member function.
typedef void (Base1::*FooPtr)(void);
FooPtr fooVTable[] = {&Base1::foo, (FooPtr)&Derived::foo};
void Base1Ptr::foo () const
{
FooPtr targetFooFunction = fooVTable[objectID];
(ptr->*targetFooFunction)();
}

// The same goes for Base2.
typedef void (Base2::*BarPtr)(void);
BarPtr barVTable[] = {&Base2::bar, (BarPtr)&Derived::bar};
void Base2Ptr::bar () const
{
BarPtr targetFooFunction = barVTable[objectID];
(ptr->*targetFooFunction)();
}

// Note that the cast of the addresses of members of Derived
// to addresses of members of Base results in UB.
// However, the results are pretty much as expected.




void invokeFoo (Base1* b) {
b->foo();
}

void invokeFooVirtually (const Base1Ptr& b) {
b.foo();
}

void invokeBarVirtually (const Base2Ptr& b) {
b.bar();
}

int main () {
Base1 b(3);
Derived d(42);

std::cout << "sizeof(Base1) == " << sizeof(Base1) << "\n";
std::cout << "sizeof(Derived) == " << sizeof(Derived) << "\n";
std::cout << "sizeof(Base1*) == " << sizeof(Base1*) << "\n";
std::cout << "sizeof(Base1Ptr) == " << sizeof(Base1Ptr) << "\n";
std::cout << "sizeof(Derived*) == " << sizeof(Derived*) << "\n";
std::cout << "sizeof(DerivedPtr) == " << sizeof(DerivedPtr) << "\n";

invokeFooVirtually(&b);
invokeFooVirtually(&d);
invokeBarVirtually(&d);
}

Regards,
Stuart

 
Reply With Quote
 
 
 
 
lieve again
Guest
Posts: n/a
 
      09-30-2012
On 26 sep, 23:39, Stuart <(E-Mail Removed)> wrote:
> On 9/23/12 "lieve again" wrote:
> [snip]
>
> > I thought maybe making the interfaces pure virtual,
> > there was a way to avoid the extra vpointers and I wanted to know how.
> > Then if I start adding pure virtual classes to impose the derived
> > classes with some kind of features like:
> > class Derived : implements Readable, Writeable, Comparable,
> > Convertible ...
> > regardless of the programming language, we are ending with instances
> > of the derived classes having 20 bytes or more even being those
> > classes with no members or empty. It is good to know.

>
> There is one way to avoid object bloat, but you are not going to like it
>
> The following code uses a hand-made vtable substitute. For that reason I
> had to introduce a special pointer type which stores the class's ID
> together with the pointer to the object. This class ID is assigned by
> hand, so that this scheme will only work for class hierarchies that will
> not get extended (if you want to use this for extendable class
> hierarchies, you'd have to use growable look-up table instead of a fixed
> array of method pointers, but then you had better use objective-C++).
>
> Note that the base class Base in my example must not contain any virtual
> methods, or else the class Derived will get bloated again. Of course,
> this makes the code really awfull to look at. However, most of it may
> get generated by a some clever pre-processor magic.
>
> Since the full code may cause some shock, I'll give a short summary:
>
> class Base {
> public:
> * * *void foo () {std::cout << "Base::foo\n";}
> * * *BasePtr operator& ();
>
> };
>
> class Derived : public Base {
> public:
> * * *void foo () {std::cout << "Derived::foo\n";}
> * * *DerivedPtr operator&();
>
> }
>
> Base1Ptr, Base2Ptr and DerivedPtr and those special pointer types that
> allow us to do the following:
>
> void invokeFooVirtually (const BasePtr& b) {
> * * *b.foo();
>
> }
>
> int main () {
> * * *Base b;
> * * *Derived d;
> * * *invokeFooVirtually(&b);
> * * *invokeFooVirtually(&d);
>
> }
>
> will print:
>
> Base1::foo
> Derived::foo
>
> There is a price you have to pay: The pointer types are now twice as
> large (the xxxPtr contains not only the raw pointer but also the class's
> ID). Furthermore the cast from DerivedPtr to BasePtr has to call the
> cast operator of DerivedPtr.
>
> A compilable example with two base classes and some members:
>
> #include <iostream>
>
> class Base1Ptr {
> protected:
> * * *static const int classID;
> * * *class Base1* ptr;
> * * *int objectID;
> public:
> * * *Base1Ptr (class Base1* ptr, int objectID = classID)
> * * * *: ptr(ptr), objectID(objectID) {}
> * * *void foo () const;};
>
> const int Base1Ptr::classID = 0;
>
> class Base1 {
> protected:
> * * *int base1Int;
> public:
> * * *Base1 (int i) : base1Int(i / 2) {}
> * * *void foo () { std::cout << "Base1::foo with base1Int = "
> * * * * * * * * * * * * * * *<< base1Int <<"\n";}
> * * *Base1Ptr operator& () {return Base1Ptr(this);}
>
> };
>
> class Base2Ptr {
> protected:
> * * *static const int classID;
> * * *class Base2* ptr;
> * * *int objectID;
> public:
> * * *Base2Ptr (class Base2* ptr, int objectID = classID)
> * * * *: ptr(ptr), objectID(objectID) {}
> * * *void bar () const;};
>
> const int Base2Ptr::classID = 0;
>
> class Base2 {
> protected:
> * * *int base2Int;
> public:
> * * *Base2(int i) : base2Int(i * 2){}
> * * *void bar () { std::cout << "Base2::bar with base2Int = "
> * * * * * * * * * * * * * * *<< base2Int <<"\n";}
> * * *Base2Ptr operator* ();
>
> };
>
> class DerivedPtr {
> * * *class Derived* ptr;
> * * *int objectID;
> public:
> * * *DerivedPtr (class Derived* ptr, int objectID)
> * * * *: ptr(ptr), objectID(objectID) {}
> * * *operator Base1Ptr();
> * * *operator Base2Ptr();
>
> };
>
> class Derived : public Base1, public Base2 {
> * * *static const int classID;
> public:
> * * *Derived (int i) : Base1(i), Base2(i){}
> * * *void foo () { std::cout << "Derived::foo with base1Int = "
> * * * * * * * * * * * * * * *<< base1Int <<"and base2Int = "
> * * * * * * * * * * * * * * *<< base2Int <<"\n";}
> * * *void bar () { std::cout << "Derived::bar\n";}
> * * *DerivedPtr operator&() {return DerivedPtr(this,classID);}
>
> };
>
> const int Derived::classID = 1; // std::max(Base1::classID,
> * * * * * * * * * * * * * * * * *// ** * * *Base2::classID) + 1;
> DerivedPtr:perator Base1Ptr() {return Base1Ptr(ptr, objectID);}
> DerivedPtr:perator Base2Ptr() {return Base2Ptr(ptr, objectID);}
>
> // Base1Ptr uses the following table to look up
> // the correct member function.
> typedef void (Base1::*FooPtr)(void);
> FooPtr fooVTable[] = {&Base1::foo, (FooPtr)&Derived::foo};
> void Base1Ptr::foo () const
> {
> * * *FooPtr targetFooFunction = fooVTable[objectID];
> * * *(ptr->*targetFooFunction)();
>
> }
>
> // The same goes for Base2.
> typedef void (Base2::*BarPtr)(void);
> BarPtr barVTable[] = {&Base2::bar, (BarPtr)&Derived::bar};
> void Base2Ptr::bar () const
> {
> * * *BarPtr targetFooFunction = barVTable[objectID];
> * * *(ptr->*targetFooFunction)();
>
> }
>
> // Note that the cast of the addresses of members of Derived
> // to addresses of members of Base results in UB.
> // However, the results are pretty much as expected.
>
> void invokeFoo (Base1* b) {
> * * *b->foo();
>
> }
>
> void invokeFooVirtually (const Base1Ptr& b) {
> * * *b.foo();
>
> }
>
> void invokeBarVirtually (const Base2Ptr& b) {
> * * *b.bar();
>
> }
>
> int main () {
> * * *Base1 b(3);
> * * *Derived d(42);
>
> * * *std::cout << "sizeof(Base1) == " << sizeof(Base1) << "\n";
> * * *std::cout << "sizeof(Derived) == " << sizeof(Derived) << "\n";
> * * *std::cout << "sizeof(Base1*) == " << sizeof(Base1*) << "\n";
> * * *std::cout << "sizeof(Base1Ptr) == " << sizeof(Base1Ptr) <<"\n";
> * * *std::cout << "sizeof(Derived*) == " << sizeof(Derived*) <<"\n";
> * * *std::cout << "sizeof(DerivedPtr) == " << sizeof(DerivedPtr) << "\n";
>
> * * *invokeFooVirtually(&b);
> * * *invokeFooVirtually(&d);
> * * *invokeBarVirtually(&d);
>
> }
>
> Regards,
> Stuart


Interesting (and complex) example. I think, I have understood it, its
a way to avoid the extra vpointers in class by make the base pointers
"fatter". Probably some programming language have implemented multiple
inheritance that way.


Looking the code, another way to do it were:

// typedef void (Base1::*FooPtr)(void);
// FooPtr fooVTable[] = {&Base1::foo, (FooPtr)&Derived::foo};
void Base1Ptr::foo () const
{
FooPtr targetFooFunction = fooVTable[objectID];
(ptr->*targetFooFunction)();

if(this->objectID != 0){
Derived* derived = static_cast<Derived*>(ptr);
derived->foo();
}else // not derived, call it normally
ptr->foo();
}

With the objectID we already know the true class of the base pointer,
so we could simply convert it safely to the derived class, in that way
we can avoid the pointer to member functions which they are a little
bit complicated. Instead we can made a table of int like an index to
know the true class or at least if we need to use virtual functions,
something like:
enum Index { BASE1 = 0, BASE2 = 1, BASE3 = 2, DERIVED = 3};

an depending of the table answer, convert it to the right class.

We could store the object id directly in the Base1, and forget about
the BasePtr and the conversion operators, but this would made the base
classes bigger, so the Base1 approach its better: we make the pointer
fatter only if an conversion from the derived class to a base class
take place.

Thanks!!
 
Reply With Quote
 
 
 
 
Cholo Lennon
Guest
Posts: n/a
 
      10-01-2012
On 09/22/2012 03:10 PM, Richard Damon wrote:
> On 9/20/12 8:59 AM, lieve again wrote:
>>
>> So, sizeof(Base1) == sizeof(Base2) == sizeof(NormalInheritance) == 4
>> bytes (only one virtual pointer)
>> but sizeof(MultipleInheritance) == 8 bytes
>> if class MultipleInheritance would inherit from another Base3, the
>> size would be 12 bytes and so on.
>>

> And this should be expected as we would expect
> sizeof(MultipleInheritance) = sizeof(Base1) + sizeof(Base2) +
> sizeof(stuff added in MultipleInheritance)
>
> as we would normally expect that each of the base classes be fully
> represented within the derived class so it is easy to treat the derived
> class as if it was any of its base classes. (There is an exception for
> empty base classes which must have a sizeof > 0 as a class by itself,
> but might not take any extra room when derived from.)
>
>> So with multiple inheritance we ends with big classes because of the
>> need of extra virtual pointers,
>> to avoid that, almost every language doesn't implement multiple
>> inheritance but Interfaces, where one
>> can only inherit more than one class but being that classes abstract
>> or pure virtual, like:
>>
>> class Base1{
>> virtual void func() = 0;
>> virtual void func3();
>> };
>>
>> class MixedClass : public NormalClass implements Base1, Base2,
>> Base...
>>
>> My question is: Don't we have the same implementation problem as in C+
>> +? Because even being these classes
>> abstract, they need a virtual pointer. Why do they impose that rule
>> in languages like C#, Java, D...?
>> Someone know the reason?
>>

>
> When inheriting from Interfaces, the difference is that the Interface
> never needs to exist as a discrete object, so there isn't a need to save
> a vtable pointer for each Interface. The Interface routine likely need a
> pointer into the base class vtable to the vtable for that interface, but
> that should be computable from the object normal vtable pointer.
>


That's why VC++ has an extension, __declspec(novtable) to mark abstract
classes intended to be used as interfaces.

> The main reason many languages don't implement multiple inheritance (but
> maybe Interfaces) is NOT object size, but language a program complexity.
>



--
Cholo Lennon
Bs.As.
ARG

 
Reply With Quote
 
Richard Damon
Guest
Posts: n/a
 
      10-01-2012
On 9/30/12 11:04 PM, Cholo Lennon wrote:
> On 09/22/2012 03:10 PM, Richard Damon wrote:
>>
>> When inheriting from Interfaces, the difference is that the Interface
>> never needs to exist as a discrete object, so there isn't a need to save
>> a vtable pointer for each Interface. The Interface routine likely need a
>> pointer into the base class vtable to the vtable for that interface, but
>> that should be computable from the object normal vtable pointer.
>>

>
> That's why VC++ has an extension, __declspec(novtable) to mark abstract
> classes intended to be used as interfaces.
>


Unfortunately, __declspec(novtable) does NOT remove the vtable from the
object, but the removes the instruction to initialize the pointer from
the constructor, and thus likely removes the vtable itself (which occurs
only once per class, not once per object, so only a small savings). The
class object will still need to have a vtablepointer in it, so members
of that class can find their virtual functions.

The "interface" class members still need to be passed a "this" object of
the appropriate type, and that object needs to have a vtablepointer in
it to find the virtual functions. The problem here is that the "inteface
class" isn't something different than a regular class.

The typical difference with real interface classes in other languages is
that those classes don't take a pointer to the interface class object,
but to the full object, they "know" that they are going to be used as a
"mixin" class. They will normally take as their calling sequence API a
pointer to the object (not their "sub object", as those language tend
not to define a thing called a "sub object", and a pointer to the
section of the vtable which acts as the vtable for interface. Since
interfaces normally can not have member variables, only functions, there
is no need to have a pointer to that.
 
Reply With Quote
 
Pavel
Guest
Posts: n/a
 
      10-07-2012
lieve again wrote:
> On 23 Sep., 23:25, Öö Tiib <(E-Mail Removed)> wrote:
>> On Sunday, 23 September 2012 23:40:42 UTC+3, lieve again wrote:
>>> Ok, so the main reason for not implementing multiple inheritance
>>> (without workarounds) is the complexity added to the programmers
>>> (learning curve) and to the compiler developers (diamond
>>> problem, ...).

>>
>> Yes, that is the reason. Also the solution, virtual inheritance is
>> not too efficient nor simple.
>>
>>> I thought maybe making the interfaces pure virtual,
>>> there was a way to avoid the extra vpointers and I wanted to know how.
>>> Then if I start adding pure virtual classes to impose the derived
>>> classes with some kind of features like:
>>> class Derived : implements Readable, Writeable, Comparable,
>>> Convertible ...
>>> regardless of the programming language, we are ending with instances
>>> of the derived classes having 20 bytes or more even being those
>>> classes with no members or empty. It is good to know.

>>
>> The bytes actually are cheap these days ... unless you write for some 8 bit
>> controller. On common platforms most of the memory goes into visuals
>> and sounds and helper texts and other massive data like that. Couple of bytes
>> for vtables of object that manages such data are usually not worth talking
>> about.
>>
>> OTOH on the 8-bit controllers where you really count bytes you do not have
>> much need for such large class hierarchy anyway.

>
> Ok, so its a problem suffered from all the actual programming
> languages, I think it could be a kind of limitation to obtain so big
> objects, but its so.
> Maybe the way to impose some kind of properties or functions to a
> class without the vpointers replication penalty is the concepts
> extension of C++11.
>
> Regards,
>

I think the above is not accurate. C++ code does suffer performance penalties
from using multiple inheritance. Moreover, and what's especially frustrating,
even the code that does not use multiple inheritance (in fact, any code using
virtual functions) suffers from at least one performance penalty imposed by the
way C++ supports multiple inheritance: the necessity to read the offset of the
call target within the object of the most-derived class overriding the virtual
method and subtracting this offset from the passed pointer to let the virtual
function implementation access to the object it expects.

Languages with single inheritance can assign a single offset from the start of
the virtual table of the most-derived class of an object to the start of the
slice of that class' virtual table correspondent to the virtual table of any of
its bases. This effectively means that any class in such a language can have a
single virtual table and the objects of the most derived class and the
correspondent objects of all its base classes can have a single address.

Complicating C++ specification by introducing special kind of classes that would
be forbidden from being used as base classes in multiple inheritance (or, even
more complex but more rewarding as well -- the classes that cannot be used as
other-than-first base classes in multiple inheritance) could eliminate this
penalty for the language users who do not use multiple inheritance.

C++ does not have a chance of assigning any virtual function once defined in a
class a single offset in a virtual table; therefore, it has to have multiple
virtual tables. As it is, that is without the complication mentioned above, C++
can not let its compiler know at a virtual call site that the call is on the
object that is the first base of the most-derived class; hence the necessity to
always read and apply the offset at run-time.

Languages with single inheritance and interfaces still have same performance
advantages as languages without interfaces for the classes that do not implement
interfaces (that is, they live to the promise of not imposing cost of a feature
on the code that does not use it better than C++ does). For a class that does
implement interfaces, the implementations have a choice of either avoiding space
cost but making calls by interface significantly more expensive or adding
pointers to individual "interface virtual tables" to the layout of an object of
such a class and having calls by an interface only one indirection more
expensive than regular virtual calls. I believe Java 1.0 took the first path and
Java 1.1 and all its further versions took the second.

-Pavel

 
Reply With Quote
 
Richard Damon
Guest
Posts: n/a
 
      10-08-2012
On 10/7/12 4:57 PM, Pavel wrote:

> I think the above is not accurate. C++ code does suffer performance
> penalties from using multiple inheritance. Moreover, and what's
> especially frustrating, even the code that does not use multiple
> inheritance (in fact, any code using virtual functions) suffers from at
> least one performance penalty imposed by the way C++ supports multiple
> inheritance: the necessity to read the offset of the call target within
> the object of the most-derived class overriding the virtual method and
> subtracting this offset from the passed pointer to let the virtual
> function implementation access to the object it expects.
>


This is incorrect. It is possible to setup the virtual table so that
virtual functions based on the 1st base class make direct calls to the
destination functions (since no pointer adjustment is needed) but if a
class is the later base, there are actually 2 tables of pointers to the
functions, one pointer to by the base sub object, and a second one, part
of the table pointed to by the 1st base object. These two different
tables point to different points, one being to a "thunk" that adjusts
the this pointer, and the other which doesn't.

Base class functions, which see the this pointer as the later base
object, only have the base pointer in that base object, this version
does not adjust the pointer if the function was last defined under that
base object, but if the function has been overridden since the multiple
inheritance, it is a thunk which adjust the this pointer and then goes
to the override.

Functions after the multiple inheritance use the pointer in the first
base class, which does the reverse, having a thunk if the function was
last overridden before the multiple inheritance, and a direct connection
if after.


Zero cost unless there is multiple inheritance, and then only for calls
to functions that do require adjusting of the this pointer.
 
Reply With Quote
 
Stuart
Guest
Posts: n/a
 
      10-08-2012

[the OP, "lieve again", observed the problem that multiple inheritance
of either interface or non-interface classes leads to object bloat due
to the necessity to stuff objects with multiple vtables]

lieve again wrote:
>> Ok, so its a problem suffered from all the actual programming
>> languages, I think it could be a kind of limitation to obtain so big
>> objects, but its so.
>> Maybe the way to impose some kind of properties or functions to a
>> class without the vpointers replication penalty is the concepts
>> extension of C++11.


On 10/7/12 Pavel wrote:
> I think the above is not accurate. C++ code does suffer performance
> penalties from using multiple inheritance. Moreover, and what's
> especially frustrating, even the code that does not use multiple
> inheritance (in fact, any code using virtual functions) suffers from at
> least one performance penalty imposed by the way C++ supports multiple
> inheritance: the necessity to read the offset of the call target


What's the call target? Never heard this term.

> within
> the object of the most-derived class overriding the virtual method and
> subtracting this offset from the passed pointer to let the virtual
> function implementation access to the object it expects.


I don't get what you mean. Can you give an example?

> Languages with single inheritance can assign a single offset from the
> start of the virtual table of the most-derived class of an object to the
> start of the slice of that class' virtual table correspondent to the
> virtual table of any of its bases.


Yeah, for C++ this offset will always be zero.

> This effectively means that any class
> in such a language can have a single virtual table and the objects of
> the most derived class and the correspondent objects of all its base
> classes can have a single address.


Right. So casting a Derived* pointer to a Base* pointer for
single-inheritance chains will always be a noop under C++. I don't see
any kind of performance penalty.

[snip]

> C++ does not have a chance of assigning any virtual function once
> defined in a class a single offset in a virtual table; therefore, it has
> to have multiple virtual tables. As it is, that is without the
> complication mentioned above, C++ can not let its compiler know at a
> virtual call site that the call is on the object that is the first base
> of the most-derived class; hence the necessity to always read and apply
> the offset at run-time.


Which offset are you talking about? Can you give an example (preferably
for the Intel architecture)?

Regards,
Stuart
 
Reply With Quote
 
Pavel
Guest
Posts: n/a
 
      10-10-2012
Stuart wrote:
>
> [the OP, "lieve again", observed the problem that multiple inheritance of either
> interface or non-interface classes leads to object bloat due to the necessity to
> stuff objects with multiple vtables]
>
> lieve again wrote:
>>> Ok, so its a problem suffered from all the actual programming
>>> languages, I think it could be a kind of limitation to obtain so big
>>> objects, but its so.
>>> Maybe the way to impose some kind of properties or functions to a
>>> class without the vpointers replication penalty is the concepts
>>> extension of C++11.

>
> On 10/7/12 Pavel wrote:
>> I think the above is not accurate. C++ code does suffer performance
>> penalties from using multiple inheritance. Moreover, and what's
>> especially frustrating, even the code that does not use multiple
>> inheritance (in fact, any code using virtual functions) suffers from at
>> least one performance penalty imposed by the way C++ supports multiple
>> inheritance: the necessity to read the offset of the call target

>
> What's the call target? Never heard this term.


The pointer to the base class on which a virtual function is called.

>
>> within
>> the object of the most-derived class overriding the virtual method and
>> subtracting this offset from the passed pointer to let the virtual
>> function implementation access to the object it expects.

>
> I don't get what you mean. Can you give an example?

Because the compiler does not know (generally, Richard gave a good algo that can
solve the issue -- but for cost) whether the base is the first base in the
particular most-derived class, it has to read the offset of the sub-object in
the object in which the virtual table is defined and subtract it from given
pointer to the base class, at run-time.

>
>> Languages with single inheritance can assign a single offset from the
>> start of the virtual table of the most-derived class of an object to the
>> start of the slice of that class' virtual table correspondent to the
>> virtual table of any of its bases.

>
> Yeah, for C++ this offset will always be zero.

For calling a virtual functions defined in a class with multiple bases by base
pointer, it will be zero only if the base is the first base (assuming without
loss of generality that the compiler allocates first base at the lowest
address); otherwise it will be something else. But, even if it is zero, it is
not known in advance to the compiler so the code will still have to read that
zero from memory and subtract it. That extra memory read can be relatively
expensive (subtraction is usually not).

>
>> This effectively means that any class
>> in such a language can have a single virtual table and the objects of
>> the most derived class and the correspondent objects of all its base
>> classes can have a single address.

>
> Right. So casting a Derived* pointer to a Base* pointer for single-inheritance
> chains will always be a noop under C++.

Right.
I don't see any kind of performance
> penalty.'

The performance penalty will is incurred to cat Base* to Derived* which is what
happens when you call Derived's overridden virtual function by pointer to Base.

>
> [snip]
>
>> C++ does not have a chance of assigning any virtual function once
>> defined in a class a single offset in a virtual table; therefore, it has
>> to have multiple virtual tables. As it is, that is without the
>> complication mentioned above, C++ can not let its compiler know at a
>> virtual call site that the call is on the object that is the first base
>> of the most-derived class; hence the necessity to always read and apply
>> the offset at run-time.

>
> Which offset are you talking about? Can you give an example (preferably for the
> Intel architecture)?


It's more compiler-specific than hardware-platform-specific. Imagine, compiler
lays out objects with virtual functions by putting virtual table pointer before
an object; and, for multiple inheritance, it places base sub-objects at the
beginning of derived objects, in the order of its base specifier list. Then for
these classes:

// file b.h
struct B { int b; virtual int getBOffset() const
{ return 0; }
};
// file b2.h
struct B2 { int b2; };
// file d.h
#include "b.h"
#include "b2.h"
struct D: public B, public B2 { int d; virtual int getBOffset() const; };
// file d.cpp
#include "d.h"
int D::getBOffset() const {
return (const char*)this - (const char*)(const B*)this;
// not zero, most likely sizeof(int)
}
// file d2.h
#include "b.h"
#include "b2.h"
struct D2: public B2, public B { int d2; virtual int getBOffset() const; };
// file d2.cpp
#include "d2.h"
int D2::getBOffset() const {
return (const char*)this - (const char*)(const B*)this;
// most likely, zero
}
// file x.cpp
#include "d.h"
#include "d2.h"
B *createB1() { return new D1(); }
B *createB2() { return new D2(); }
// file client.cpp
#include "b.h"
#include "x.h"
B *createB1();
B *createB2();
B *bPtr = createB1();
int o1 = bPtr->getBOffset();
B *bPtr2 = createB2();
int o2 = bPtr2->getBOffset();

Above, we know that to call bPtr2->getBOffset() (which is actually
D2::getBOffset()) compiler does not have to subtract anything from *bPtr2, but
compiler does not (D and D2 are not visible in client.cpp and neither are the
definitions of createB1 or createB2. Therefore, compiler has to generate th code
that reaches for the virtual table of bPtr2 and retrieve that zero (sometimes
stored at some negative address in the virtual table and sometimes in other
ways). As for the clever trick with double-virtual table and thunks described by
Richard Damon to avoid that, it will work but it comes at some cost for calling
virtual functions on non-second base and thus is not always employed (I will
stop at it later in the answer to his post).

>
> Regards,
> Stuart


HTH
-Pavel
 
Reply With Quote
 
Pavel
Guest
Posts: n/a
 
      10-10-2012
Richard Damon wrote:
> On 10/7/12 4:57 PM, Pavel wrote:
>
>> I think the above is not accurate. C++ code does suffer performance
>> penalties from using multiple inheritance. Moreover, and what's
>> especially frustrating, even the code that does not use multiple
>> inheritance (in fact, any code using virtual functions) suffers from at
>> least one performance penalty imposed by the way C++ supports multiple
>> inheritance: the necessity to read the offset of the call target within
>> the object of the most-derived class overriding the virtual method and
>> subtracting this offset from the passed pointer to let the virtual
>> function implementation access to the object it expects.
>>

>
> This is incorrect. It is possible to setup the virtual table so that
> virtual functions based on the 1st base class make direct calls to the
> destination functions (since no pointer adjustment is needed) but if a
> class is the later base, there are actually 2 tables of pointers to the
> functions, one pointer to by the base sub object, and a second one, part
> of the table pointed to by the 1st base object. These two different
> tables point to different points, one being to a "thunk" that adjusts
> the this pointer, and the other which doesn't.
>
> Base class functions, which see the this pointer as the later base
> object, only have the base pointer in that base object, this version
> does not adjust the pointer if the function was last defined under that
> base object, but if the function has been overridden since the multiple
> inheritance, it is a thunk which adjust the this pointer and then goes
> to the override.
>
> Functions after the multiple inheritance use the pointer in the first
> base class, which does the reverse, having a thunk if the function was
> last overridden before the multiple inheritance, and a direct connection
> if after.
>
>
> Zero cost unless there is multiple inheritance, and then only for calls
> to functions that do require adjusting of the this pointer.
>

True but thunk approach comes at higher cost for calling at non-first base of
extra jump as compared with 'classic' approach. Extra jump in chunk is
equivalent to at least one extra memory read (only to instruction instead of
data cache) and some instruction decoding. The approach also somewhat increases
memory usage with the second virtual table and thunks. But you are right about
zero-cost if calls by non-first base are not used -- I completely forgot about
thunk approach.

-Pavel
 
Reply With Quote
 
Stuart
Guest
Posts: n/a
 
      10-10-2012
>> On 10/7/12 Pavel wrote:
>>> C++ code does suffer performance
>>> penalties from using multiple inheritance. Moreover, and what's
>>> especially frustrating, even the code that does not use multiple
>>> inheritance (in fact, any code using virtual functions) suffers from at
>>> least one performance penalty imposed by the way C++ supports multiple
>>> inheritance: the necessity to read the offset of the call target within
>>> the object of the most-derived class overriding the virtual method and
>>> subtracting this offset from the passed pointer to let the virtual
>>> function implementation access to the object it expects.


I still don't agree. If you use single-inheritance under C++, there is
no performance penalty compared to any vtable-based programming language
which would result from the fact that C++ provides multiple-inheritance.
Since the vtables and the data members of classes can be laid out by the
compiler in such a fashion that any "this" pointer never has to be
adjusted in a single-inheritance hierarchy, C++ is as effecient as it
ever gets.

[snip]
> Because the compiler does not know whether the base is the first
> base in the particular most-derived class, ... [snip]


If the compiler wants to do anything with a class, it needs to know the
complete definition of the class and all its base classes. So it knows
whether a certain base class is the first base class of another class or
not. What you probably tried to say is that the compiler, receiving a
pointer to Base* cannot know whether the real type of the object is Base
or Derived, so a "this"-pointer adjustment may have to be performed when
any of Derived's methods should be invoked.

Note that this adjustment is most likely done inside a "thunk"-method
which simply adjusts the "this" pointer and invokes Derived's
implementation method that does not need to adjust the "this" pointer.

Like so:

class Base1 {
int b1;
virtual void print1 ();
};
class Base2 {
int b2;
virtual void print2 ();
};
class Derived : Base1, Base2 {
int derived;
virtual void print1 ();
virtual void print2 ();
};

,____________________ ___________ ,________________________
+0 | |vtable |->. print1 | |void Derived:rint2 { |
|Base1 |----------| | print2 |->| std::cout << derived;|
+4 | |int b1; | |_________| |//&derived == this+16; |
|________|__________| ___________ |} |
+8 | |vtable |->| print2 | |_______________________|
|Base2 |----------| |_________| ,________________________
+12 | |int b2; | | |__asm { |
|________|__________| |---->| this -= 8; |
+16 |int derived; | | call Derived:rint2 |
|___________________| |} |
|_______________________|

And even that is an implementation detail: A compiler writer may
choose to generate Derived:rint multiple times, each version using its
own set of offsets. This would eliminate the need for thunking code
completely at the cost of larger executables:
,____________________ ___________ ,________________________
+0 | |vtable |->. print1 | |void Derived:rint2 { |
|Base1 |----------| | print2 |->| std::cout << derived;|
+4 | |int b1; | |_________| |//&derived == this+16; |
|________|__________| ___________ |} |
+8 | |vtable |->| print2 | |_______________________|
|Base2 |----------| |_________|
+12 | |int b2; | | ,________________________
|________|__________| |---->|void Derived:rint2 { |
+16 |int derived; | | std::cout << derived;|
|___________________| |//&derived == this+8; |
|} |
|_______________________|

Regards,
Stuart
 
Reply With Quote
 
 
 
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
inheritance, multiple inheritance and the weaklist and instance dictionaries Rouslan Korneychuk Python 8 02-10-2011 04:02 AM
Interface inheritance vs Implementation inheritance. Daniel Pitts Java 27 02-27-2008 01:37 AM
Multiple inheritance/interface delegate through template function not working roman.blackhammer@gmail.com C++ 7 07-03-2007 09:00 AM
Multiple inheritance: Interface problem workaround, please comment this Axel Straschil Python 6 04-11-2005 08:14 AM
Private access modifier and Inheritance (Inheritance implementation in Java) maxw_cc Java 1 12-21-2003 11:38 AM



Advertisments