Inline limitation?

[Joe Buck]

> There is such a warning, -Winline. I am not a compiler guru so I don't
> know how difficult it is to do inline regardless of the order of
> definition and use of functions. I know that our Sun C++ Version 4.2
> does inlining regardless of the order of definition.

No, Sun's compiler is not omniscient and can't always inline regardless of
the order of definition (though it evidently can inline in a couple of
cases where g++ cannot).

> This is very helpful in C++ because there are a lot of short,
> inlinable functions but the order of #includes forbids a
> define-before-use ordering. So egcs is losing a big opportunity
> to optimize here.

I doubt it.    You may be forming the mistaken impression that g++ can
never inline a function that appears after its call, however in at least
some cases it can (e.g for functions defined in the class declaration, to
a limited degree).

If you are worried that you are losing optimization, turn on -Winline,
which will tell you which functions cannot be inlined.  You could then try
reordering only the most important calls.