https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57955

--- Comment #32 from Richard Biener <rguenth at gcc dot gnu.org> ---
So we do not "vectorize" this because

  x = *.LC0;

is an aggregate copy the vectorizer doesn't handle.  I was specifically
looking for the *.LC0 load to be constant folded so we'd end up with
another constant pool entry for the vector constant, properly aligned
this time.  Eventually we run this through by-pieces which in turn
might or not might be to perform constant folding.  It might even
increase the alignment of *.LC0 in case it is not emitted yet.

Reply via email to