https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117182
--- Comment #13 from Oleg Endo <olegendo at gcc dot gnu.org> --- (In reply to Vladimir Makarov from comment #12) > > Simply I found a wrong (with my POV) transformation in the testcase and the > patch fixes it. And the patch has the same effect (but not in all cases) as > Kazumoto's patch for cannot_substitute_const_equiv_p. > Hm, OK. You know best which transformation is proper or not in that place. Of course you can add it, but it doesn't fix the issue at hand. LRA will probably need to add some calls to validate insn changes.
