[Bug ipa/103227] [12 Regression] 58% exchange2 regression with -Ofast -march=native on zen3 since r12-5223-gecdf414bd89e6ba251f6b3f494407139b4dbae0e

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103227

--- Comment #14 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Seems the performance is now better than before
https://lnt.opensuse.org/db_default/v4/SPEC/graph?highlight_run=21683&plot.0=286.407.0
Still I think I should implement the logic to stabilize the order of nodes and
edges in callgraph and lets discuss if we can make inline heuristics more
robust here.