Re: Volatile MEMs in statement expressions and functions inlined astrees

[Linus Torvalds <torvalds at transmeta dot com>]

On 14 Dec 2001, Alexandre Oliva wrote:
>
> > See? We end up reloading 'a' - simply because on the left side of the
> > assignment we have to use the lvalue.
>
> Interesting.  That's a very reasonable and consistent approach.  But

There's always a "but.." ;)

>   ``The behavior of an expression of the form E1 op= E2 is equivalent
>   to E1=E1 op E2 except that E1 is evaluated only once.'' [expr.ass]/7
>
> seems to imply that:
>
>                 (a += b) += c;
>
> should be equivalent to
>
>                 a = (a += b) + c;

The way I parse it, E1 is "(a += b)" and E2 is "c", with "op" being "+".

Right?

So the correct expansion of "E1 op= E2" to "E1 = E1 op E2" is in fact

	(a += b) = (a += b) + c;

together with the "evaluate only once" clause. Agreed? That is _not_
really equivalent to your expansion.

But to be quite honest, I have no idea _what_ it is equivalent to, as that
expression just makes me go "Huh!?" ;)

My argument hinges on the fact that I think evaluating "(a += b)" returns
two different things, both an rvalue _and_ an lvalue, and which of them is
"E1"? E1 should be the _same_ on both sides, and on the left side it has
to be a lvalue, so you can argue that the right side should also use the
lvalue, and that would make "(a += b) = (a += b) + c;" (with the "only
once") really become

	a += b;
	a = a + c;		// reload 'a'

while if you say that E1 is the lvalue on the left side, and a different
entity (the rvalue) on the right side, then you do indeed get

	a += b;			// "tmp" = rvalue, ie the value of "a+b".
	a = tmp + c;		// no reload.

So I think it depends on what you think E1 is. Does it mean

	E1 = (lvalue) (a+=b)

	E1 = E1 + c

or does it mean

	E1.left = (lvalue) (a+=b)
	E1.right = (rvalue) (a+=b)

	E1.left = E1.right + c;

I vote for the first one, but I'll admit that that is mainly because it
gets the behaviour I want ;)

		Linus