https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121606
--- Comment #7 from Hongtao Liu <liuhongt at gcc dot gnu.org> --- (In reply to rguent...@suse.de from comment #5) > On Wed, 20 Aug 2025, haochen.jiang at intel dot com wrote: > > > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121606 > > > > --- Comment #4 from Haochen Jiang <haochen.jiang at intel dot com> --- > > (In reply to Richard Biener from comment #3) > > > Hmm, but why would __attribute__((target("avx10.1"))) care for global > > > -mno-*? > > > > It is caused by we are keeping that behavior same as command line option and > > going through the whole ix86_option_override_internal, resulting in it is > > equivalent to "-march=native -mavx10.1". > > > > In this case, it will finally become "-march=znver5 -mxyz -mno-abc > > -mavx10.1", > > where xyz are the features Zen5 having, abc are those Zen5 not having. > > -mno-avx512fp16 -mavx10.1 leads to the warning. > > But > > ./cc1 -quiet t.c -march=znver5 -mno-avx512fp16 -mavx10.1 > > doesn't diagnose it. And I wouldn't have expected a diagnostic for this > as -mavx10.1 should simply re-enable avx512fp16? Are you using GCC16, and yes GCC16 is refactored with that logic.
