https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63968
--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> --- Problem was introduced by adding gcc_assert that guards if we really decrease a key. I'm testing patch that implements slow path: if a key is bigger than current value, we process correct increase operation (deletion + insertion). Old implementation suppress replace operation if a key is bigger and there's no notification about that. bb-reorder output for 175.vpr: lChanging key for bb 5 from -155 to -1015655. Changing key for bb 12 from -533 to -1049033. Changing key for bb 19 from -533 to -1049033. Changing key for bb 22 from -53 to -1004853. Changing key for bb 31 from -1250 to -1126250. Changing key for bb 27 from -533 to -1006133. Changing key for bb 23 from -53 to -1004853. Changing key for bb 50 from -533 to -1049033. Changing key for bb 51 from -53 to -1004853. Changing key for bb 58 from -533 to -1049033. Changing key for bb 61 from -53 to -1004853. Changing key for bb 197 from -27 to -1002727. Changing key for bb 205 from -9 to -1000809. Changing key for bb 70 from -37 to -1003737. Changing key for bb 79 from -37 to -1003737. Changing key for bb 98 from -1003468 to -1001768. Thanks, Martin
