https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84419
--- Comment #4 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> --- Created attachment 43456 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=43456&action=edit Possible patch I think the problem was that the optimisation was dropping the alignment information, so we ended up using an aligned rather than an unaligned access. I don't have access to skylake hardware, but could you try the attached patch?