Re: [Xen-devel] code question?

[David Hopwood <david.nospam.hopwood@xxxxxxxxxxxxxxxx>]

Keir Fraser wrote:
> On 11 Aug 2005, at 18:03, M.A. Williamson wrote:
>
> > I think that:
> > !!a & !b === (a != 0 ) && ( b == 0 )
> >
> > It's just more efficient to do it directly using bitwise ops than to 
> > use the full logical operators (can gcc really not optimise this sort 
> > of thing???).
> It can't optimise out short-circuit semantics.
Yes it can, if the second branch of the short-circuit is guaranteed to
terminate with no side effects when it has defined behaviour.

The function thread_jump in
<http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cfgcleanup.c?rev=1.140.2.7>
attempts to prove that a basic block satisfies this property, although I'm
not sure whether gcc is capable of doing the specific optimization we're
talking about here. I know that some other C compilers are.

--
David Hopwood <david.nospam.hopwood@xxxxxxxxxxxxxxxx>


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel