------- Comment #5 from ebotcazou at gcc dot gnu dot org 2008-11-28 15:59 ------- > Even when that is fixed (can look at it), I guess there still be > possibilities that tree optimizers miss something that the RTL optimizers > might DCE away (otherwise, why would we run RTL DCE at all?).
My gut feeling is also that most dead calls are DCEd at the tree level nowadays so we could decide to restrict RTL DCE to non-call insns. > I guess option 3 is remove this i386.c optimization, assume only very few > calls will be DCEd at RTL level and count with the fact that their argument > stores might not have been eliminated. Other back-ends may do something equivalent though. I guess that option 4 is to investigate why DSE doesn't remove the dead stores. Since there are no calls in the function anymore, this should be doable. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38245
