https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81245
--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Richard Biener from comment #6) > Not sure if it's good to introduce copysign here anyway (do we actually > vectorize this w/o target support?) We should be able to without target support; I don't know if we do though. For -ffast-math we should able to convert copysign into a veccond or bit selection. I just know aarch64 has the target support already and that is why I saw the ice.
