On 10 May 2011 13:51, Marc Glisse <marc.glisse@xxxxxxxx> wrote: > On Tue, 10 May 2011, Igor Bukanov wrote: > >> is it possible to tell GCC that a pointer cannot be NULL at a >> particular point so the compiler can optimize the control flaw better >> and avoid redundant NULL checks on the fast paths? > > I seem to remember that dereferencing the pointer had that effect. Yes, this is indeed the case in the cases I checked. So I would need to refactor much less code, thanks!
