Allow vector size 8 in VectorizeLoadsAndStoresPass.
When we unroll loops, we sometimes unroll by a factor of 8. Then we will also
compute a vector size of 8, so we should allow it, otherwise we need to rely on
LLVM vectorization which may not happen in case one of the parameters is
in-place.
Allow vector size 8 in VectorizeLoadsAndStoresPass.
When we unroll loops, we sometimes unroll by a factor of 8. Then we will also compute a vector size of 8, so we should allow it, otherwise we need to rely on LLVM vectorization which may not happen in case one of the parameters is in-place.