https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89497
--- Comment #28 from Jakub Jelinek <jakub at gcc dot gnu.org> --- Yeah, I've also found that it doesn't fail on x86_64 with r269301 even when the optimized dump is identical to s390x. Debugging now, in the third iteration of the outer loop the inner loop doesn't stop where it should. Will file a latent PR once debugged.