In .NET, what is the internal implementation of a delegate?

Question

I understand that a declaration of a delegate is something like this:

public delegate int PerformCalculation(int x, int y);

However, there must be more going on. The purpose of the delegate is to provide a pointer to a method, and to do that you encapsulate the reference to the method in the delegate.

What kind of structure is this reference held in (internally in the delegate)? I also understand that you can encapsulate a reference to multiple methods in a delegate. Does this mean that there is an array in the delegate that holds these?

Also, what methods are defined in the delegate, etc. What is really happening when you declare a delegate with the terse:

public delegate int PerformCalculation(int x, int y);

?

EDIT: Some clarification. When you declare a delegate, the compiler automatically creates a sealed class that inherits from System.MulticastDelegate for you. You can see this if you look at your assembly with ildasm. This neat. Basically, with one statement, you are getting an entire new class generated for you at compile time, and it has all the functionality you need.

Answer 1

All delegates inherit from the System.Delegate type which hold a Target and Method. More precisely they inherit from System.MultiCastDelegate which inherits from System.Delegate.

Answer 2

Internally it's a reference type, quite similar to a class. Transcribed it looks like this:

public /* delegate */ class PerformCalculation : MulticastDelegate {
    public PerformCalculation(object target, IntPtr method) {}
    public virtual void Invoke(int x, int y) {}
    public virtual IAsyncResult BeginInvoke(int x, int y, AsyncCallback callback, object state) {}
    public virtual void EndInvoke(IAsyncResult result) {}
}

I left the implementations of these members empty, they are actually mapped to code in the CLR. The compiler dynamically generates the method signatures, depending on the signature of the delegate declaration. Note the x and y arguments. The JIT compiler helps to get the constructor called, using the += or -= syntax, it knows the memory address of the delegate target method. The compiler automatically generated the target argument value, depending on whether the target method was static or not. The two arguments map to the (Multicast)Delegate.Target and Method properties. The actual base class instance can be either Delegate or MulticastDelegate, depending on how many targets were subscribed.

Lots of secret sauce going on here.