https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124362
--- Comment #9 from Jeffrey A. Law <law at gcc dot gnu.org> --- I'm not really GPU debugging at this point. My hope was that there were very few deltas introduced by that DOM change and that one would stand out as obviously wrong when I started looking at the dumps. I certainly see things in the dump that are encouraging and expected in terms of validating the DOM patch is finding the kinds of things it should find. But it does create one equivalence that looks concerning to me. That one concerning equivalence cascades into a second equivalance resulting in removal of a PHI because we have two PHIs that have the same RHS in the same block. That in appears to feed into a memory store. I haven't chased the control flow beyond that as I'm more concerned about proving the equivalences are correct. I'm on a tight schedule for an unrelated project that's just about wrapped up which will allow me to focus more on the BZ shortly.
