https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127106
--- Comment #3 from Drea Pinski <pinskia at gcc dot gnu.org> --- So I think I know what is happening. since is_gimple_reg_type is true for v4u16, but IPA-SRA seemly does check if the use was inside a BFR on the lhs so it does not mark it as a DECL_NOT_GIMPLE_REG_P . I wonder if this is a latent bug for a while just exposed by some heurstic changes ...
