https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122853
--- Comment #6 from Robin Dapp <rdapp at gcc dot gnu.org> --- Thanks. I don't understand it at all, though ;) The bad asm you sent is different from what I get when compiling (and from godbolt). It even contains multiplications where we should have left shifts by 16 and widening adds. Maybe that specific commit (range) was broken? We change reduction stuff around that time so maybe that? I'll go back to that commit and check later but for now I don't see much to be done.
