https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123207
--- Comment #10 from Haochen Jiang <haochen.jiang at intel dot com> --- It seems to me that there are two more issues with SPEC on this patch. 511.povary_r with "-Ofast -march=native -flto" will build fail on many platforms (at least for Granite Rapids, Sapphire Rapids, Ice Lake, many Zen machines and AWS Graviton3/4). It seems to me that it has fixed on trunk on SPR, but I have not validated on all the machines. 523.xalancbmk_r with "-Ofast -march=native -flto" got a ~10% codesize on Granite Rapids, Sapphire Rapids, Ice Lake and AWS Graviton 3/4 after this patch. And also a ~3% perf regression on all mentioned platforms.
