https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118727
--- Comment #18 from Xi Ruoyao <xry111 at gcc dot gnu.org> --- (In reply to Xi Ruoyao from comment #14) > Anyway it does not affect AArch64, so removing it from target list. > > LoongArch is the only target having both uabd and sabd but not > vect_widen_uabd/vec_widen_sabd. > > RISC-V and PowerPC only have uabd (no sabd) and they don't have > vect_widen_uabd/vec_widen_sabd as well, but I'm not sure if they are > affected. To trigger it we must have both uabd and sabd, so an ABD with signed inputs can be recognized first, and then it can be incorrectly turned into a SAD with unsigned inputs. So having only uabd cannot trigger the issue. Thus LoongArch seems the only affected target.
