https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122096
--- Comment #3 from Haochen Jiang <haochen.jiang at intel dot com> --- (In reply to Andrew Pinski from comment #1) > This is expected I think. As GCC is (speculatively) devirtualizing some > function pointers some more and then inlining more afterwards. > I see. I am going to have a quick look into the code change. > I am not sure what more should be done here since -O3/-Ofast is > defined/documtned as it might increase the code size without any gain over > -O2. It is not a comparison to -O2 here, but -Ofast itself.
