[llvm] [LV] Add support for widening loads/stores to a VF multiple (PR #217670)

[Benjamin Maxwell via llvm-commits]

================
@@ -288,6 +291,62 @@ void UnrollState::unrollHeaderPHIByUF(VPHeaderPHIRecipe *R,
   }
 }
 
+void UnrollState::unrollMemOpWithVFMultiple(VPRecipeBase &R,
----------------
MacDue wrote:

No, it's because for the masked version of this transform we also have to rewrite the active lane mask (and insert/move some `extract-vector-part` operations). This PR only adds the unmasked support to simplify the initial transform. 

https://github.com/llvm/llvm-project/pull/217670