> On September 26, 2020 12:04:24 AM GMT+02:00, Jan Hubicka <hubi...@ucw.cz> > wrote: > >Hi, > >while adding check for gimple_clobber I reversed the return value > >so instead of ignoring the statement ipa-modref gives up. Fixed thus. > >This explains the drop between originally reported disambinguations > >stats and ones I got later. > > I don't think you can ignore clobbers. They are barriers for code motion.
Hmm, you are right. In pure-const we do if ((ipa || cfun->after_inlining) && gimple_clobber_p (stmt)) that I think is safe. We don't do code motion in early passes but if we will and later inline, clobber will lead to wrong code. After inlining we will never see the clobber and thus ignoring it should be safe. Honza
