Question

In java bytecode why is the receiver is pushed onto the stack first followed by all the parameters? I seem to remember it being something to do with efficiency.

This is true for both method calls and setting fields.

Method Call

class X {

    int p(int a) {
        //Do something
    }
    int main() {
        int ret = p(1);
    }

}

Main method compiles to:

aload_0 // Load this onto the stack
iconst_1 // Load constant 1 onto the stack
invokevirtual <int p(int)> from class X

Setting a field:

class X {
    int x;
    int main() {
        x = 1;
    }

}

Main method compiles to:

aload_0 // Load this onto the stack
iconst_1 // Load constant 1 onto the stack
putfield <int x> from class X

Answer 1

Being pushed first has advantages in that

• The target method can use the denser “aload0” bytecode (smaller than
  if it were much later in the parameter list and had to use a
  parametric version of the aload bytecode. Because “this” is often referenced in methods for both field and method access, it leads to a real code density improvement.
• One often does cascading method sends like “foo.bar().baz()”. When bar() returns,
  the future “this” for .baz() is magically already in the right place on the
  stack if you arrange things as was done in Java.