https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114057
Filip Kastl <pheeck at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Summary|[14 Regression] 435.gromacs |[14 Regression] 435.gromacs |fails verification with |fails verification with |-Ofast -march=znver4 and |-Ofast |PGO after |-march={znver2,znver4} and |r14-7272-g57f611604e8bab |PGO after | |r14-7272-g57f611604e8bab --- Comment #6 from Filip Kastl <pheeck at gcc dot gnu.org> --- I just observed the same miscomparison on a Zen2 machine. Exactly the same values: 0002: 3.07684e+02 3.11606e+02 So I assume it's the same problem.