https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86541
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2018-07-17
CC| |rguenth at gcc dot gnu.org
Ever confirmed|0 |1
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed. So you are asking for GCC to properly "lower" explicit source level
vector_size(8) operations to SSE (tree-vect-generic.c), not the autovectorizer
supporting this?
tree-vect-generic.c currently supports the reverse - if you'd use
vector_size(16)
then targets with only smaller vectors get those split up appropriately. That
sounds easier in case vectorization with the larger vector size is possible
for the code in question.
Given that we have a target pass that makes use of SSE regs for scalar
operations I wonder if it would make more sense to attack this at the
target level by claiming native support for vector_size(8) and using
a target pass to make that work. As you said the most simple way is to
movlhps %xmmN, %xmmN at strategic places. That very thing could be
also done by tree-vect-generic.c of course.
