Re: typedefs and structs

[Vadim Lobanov]

On Wed, 9 Nov 2005, Steven Rostedt wrote:

> On Wed, 2005-11-09 at 15:40 -0800, Vadim Lobanov wrote:
> > On Thu, 10 Nov 2005, Andreas Schwab wrote:
> >
> > > Vadim Lobanov <vlobanov@xxxxxxxxxxxxx> writes:
> > >
> > > > However, if the code is as follows:
> > > > 	void foo (void) {
> > > > 		int myvar = 0;
> > > > 		printf("%d\n", myvar);
> > > > 		bar(&myvar);
> > > > 		printf("%d\n", myvar);
> > > > 	}
> > > > If bar is declared in _another_ file as
> > > > 	void bar (const int * var);
> > > > then I think the compiler can validly cache the value of 'myvar' for the
> > > > second printf without re-reading it. Correct/incorrect?
> > >
> > > Incorrect. bar() may cast away const.  In C const does not mean readonly.
> >
> > In that case, I stand corrected.
> >
> > Is there any real reason to apply const to pointer targets, aside from
> > giving yourself a warning in the case you try to write the pointer
> > target directly? Seems to be a missed opportunity for optimizations
> > where the coder designates that it's okay to do so.
>
> Actually, where are you going to cache it? In a register? but calling
> bar() may use that register, so it would be stored on the stack anyway.

May, but not necessarily will.

> I doubt that this is a problem with the compiler, since if bar _is_
> small, then myvar is most likely already in the processor's cache to
> begin with, so it wouldn't need to go back out to memory, unless it was
> modified.

You're right, however. There's very few cases where such an optimization
would be useful, due to register constraints.

> -- Steve
>
>

-Vadim Lobanov
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/