https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122583

--- Comment #5 from Robin Dapp <rdapp at gcc dot gnu.org> ---
vect-align-2.c uses packed access and gcn does not allow that in
gcn_vectorize_support_vector_misalignment.  As mentioned, I'd return true and
try to avoid the scalar subregs somewhere else.

Reply via email to