https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122170
--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> --- Created attachment 63509 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=63509&action=edit Patch which I will test on both aarch64 and x86_64 Note this fixes the regression. I think it is always correct as optab for the code/mode if implemented will always be a direct one. Though I wonder if we should just always do code_to_optab/optab_handler for ARITHMETIC_P/UNARY_P instead of that switch statement.
