https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81363
--- Comment #2 from Bill Schmidt <wschmidt at gcc dot gnu.org> --- (In reply to Jakub Jelinek from comment #1) > This went away with r250295 which reverted some earlier commit. Was the > patch reverted because of this LRA ICE, or for some other reason. I.e. can > we close it as RESOLVED/FIXED, or is there a plan to reapply the patch > eventually and something we should get RA folks involved in? I don't think this problem matches the reason for the revert directly, but it might have caused it indirectly. Carl's original code was fine, but he added some standard patterns we had been missing before, allowing vectorization to kick in for cases where it previously hadn't. This exposed a bug in the powerpc back end, which I fixed in https://gcc.gnu.org/r251161. Carl then re-committed his patches in https://gcc.gnu.org/r252027. I don't believe the bug that I fixed could have caused the mismatched vector modes that showed up in this bug report, but I could be wrong. It's theoretically possible that there is an RA bug hiding here, but it doesn't reproduce anymore. At this point I'd suggest closing this as RESOLVED/FIXED since the work on vec_mule and vec_mulo is complete, and the bug is no longer occurring.
