It is not always true though -- only when the array address is picked by the pass to be the induction variable.
David On Thu, Nov 1, 2012 at 2:24 PM, Jakub Jelinek <ja...@redhat.com> wrote: > On Thu, Nov 01, 2012 at 02:19:43PM -0700, Xinliang David Li wrote: >> Right -- I found the same thing, it is the ivopt that resets it -- the >> IV opt pass should have a TODO_update_address_taken. > > That wouldn't change anything. The IVopts pass adds some_ptr = &a[0]; > to the IL, so a is then address taken. > > Jakub
